Replace all sources with Unpaywall/ImpactStory, which already combines all the necessary/good ones.

Remove sources that may suffer from conflict of interest: Scopus.

I think BASE is already getting good data about this question, although it may all come from crossref (?), may be worth to have a look at scienceopen data, too

Elsevier products (Scopus) have pretty obvious conflicts of interest?

Again, if you consider only Scopus - with all the proprietary/conflict of interest limitations already noticed - all these indicators should be named not "publications" but "journal articles" percentages.